AbstractObservationList: AbstractObservationList

In HopkinsIDD/ForecastFramework: A Basis for Modular Model Creation

Description

A class for storing data in multiple interrelated tables. Each table relates to the other tables through keys.

Fields

aggregate

A function used to combine covariates of the same key/val pair.

arr

This is the full array. For extensibility, it cannot be written to directly and must be modified through methods.

cellData

A list of metadata associated with the cells of the data.

cnames

The names of columns in the data.

colData

A list of metadata associated with the columns of the data.

dimData

The data associated with each dimension of the array.

dims

The size of the array.

dnames

The size of the array.

frame

Long form data.

mat

This is the matrix. For extensibility, it cannot be written to directly and must be modified through methods.

metaData

Any data not part of the main data structure.

ncol

The number of columns in the data.

ndim

The number of dimensions of the array.

nrow

The number of rows in the data

rnames

The names of rows in the data.

rowData

A list of metadata associated with the rows of the data.

Methods

debug(string)

A function for debugging the methods of this class. It calls the browser command. In order for methods to opt into to debugging, they need to implement the following code at the beginning: if(<method_name> %in% private$.debug){browser()}. This method exists, because the debugger is not always intuitive when it comes to debugging R6 methods.

Arguments

string	-	The name(s) of methods to debug as a character vector

formArray(val,...,metaData=list(),dimData=list())

This method must be extended. In order to use an ObservationList as an ArrayData, you need to select which columns to use to form the dimensions of the array. Optionally, you can also assign some of the columns to be associated with each dimension (or cell). Note that aggregate is used to determine how to deal with multiple observations associated with a particular grouping.

Arguments

val	-	The attribute of frame to use for the values of the array (must aggregate_ to a numeric type)
...	-	Column names of columns which, in order should form the dimensions of the array
metaData	-	The attribute(s) of frame to store in metaData so they can be accessed by methods expecting a MatrixData object.

initialize(...)

This function should be extended. Create a new instance of this class.

Arguments

...	-	This function should take in any arguments just in case.

undebug(string)

A function for ceasing to debug methods. Normally a method will call the browser command every time it is run. This command will stop it from doing so.

Arguments

string	-	The name(s) of the methods to stop debugging.

See Also

Inherits from : FrameData

Examples

require(dplyr)
ObservationList<- R6Class(
  classname= "ObservationList",
  inherit = AbstractObservationList,
  private = list(
    .aggregate = NULL,
    .frame = data_frame(),
    aCurrent = FALSE,
    .aDims = list(),
    .aVal = '',
    .slice = 1,
    .aDimData = list(),
    na.rm = FALSE,
    updateArray = function(na.rm = private$na.rm){
      if('updateArray' %in% private$.debug){
        browser()
      }

      if(length(self$aDims) == 0){
        warning("Form array with formArray first")
        private$.arr = array(as.numeric(NA),c(0,0))
        private$.ndim = 2
        private$.dims = c(0,0)
        private$.dimData = list(list(),list())
        private$.dnames = list()
        return()
      }
      if(nrow(private$.frame) == 0){
        warning("The frame is empty.")
        private$.arr = array(as.numeric(NA),c(1:self$aDims*0))
        private$.ndim = length(self$aDims)
        private$.dims = 1:length(self$aDims)*0
        private$.dimData = lapply(self$aDims,list())
        private$.dnames = list()
        return()
      }
      private$.dnames = lapply(self$aDims,function(name){
        as.character(unique(private$.frame[[name]]))
      })
      private$.dims = sapply(private$.dnames,length)
      private$.ndim = length(private$.dims)
      private$.arr <- private$.frame %>%
        select_(
          .dots = setNames(
            c(unlist(self$aDims),self$aVal),
            nm=c(unlist(self$aDims),self$aVal)
          )
        ) %>%
        group_by_(.dots=setNames(self$aDims,NULL)) %>%
        self$aggregate() %>%
        ungroup() %>%
        acast(as.formula(paste(self$aDims,collapse='~')),value.var=self$aVal)
      mode(private$.arr) = 'numeric'
      private$.dnames = dimnames(private$.arr)
      if(class(self$aDimData) == 'list'){
        if(length(self$aDimData) < length(self$aDims)){
          self$aDimData[length(self$aDims)] = list(NULL)
        }
      } else {
        stop("array metadata should be a list of columns in the data frame.")
      }
      private$.dimData = mapply(dim=self$aDims,data=self$aDimData,function(dim,data){
        self$frame %>%
          group_by_(dim) %>%
          self$aggregate() %>%
          ungroup() %>%
          select_(.dots=data) %>%
          as.list()
      })
      if(length(private$.dimData) < private$.ndim){
        private$.dimData = lapply(1:private$.ndim,function(x){
          if(length(private$.dimData >= x)){
            private$.dimData[[x]]
          } else {
            list()
          }
        })
      }
      private$aCurrent <- TRUE
    }
  ),
  public = list(
    initialize = function(data=data_frame(),...){
      self$frame <- as_data_frame(data)
      self$formArray(...)
      private$.aggregate = function(input_data){
        if('aggregate' %in% private$.debug){
          browser()
        }
        grouping = groups(input_data)
        input_data %>%
          input_data <- summarize_all(funs(
            sum = if(is.numeric(.) || is.logical(.)){
              sum(.,na.rm=private$na.rm)
            } else{NA},
            unique = if(length(unique(.))==1){unique(.)} else{list(unique(.))}))
        input_data <- input_data %>%
          group_by_(.dots=grouping)
        if('sum' %in% names(input_data)){
          input_data <- input_data %>%
            rename_(.dots=setNames('sum',paste(private$.aVal,'_sum',sep='')))
        }
        if('unique' %in% names(input_data)){
          input_data <- input_data %>%
            rename_(
              .dots=setNames('unique',paste(private$.aVal,'_unique',sep=''))
            )
        }
        column_names = c(
          unlist(
            sapply(private$.aDimData,function(x){paste(x,"_unique",sep='')})
          ),
          paste(private$.aVal,'_sum',sep=''))
        column_names = column_names[column_names %in% names(input_data)]
        input_data %>%
          select_(.dots=column_names) ->
          input_data
        input_data %>%
          rename_(.dots=setNames(
            names(input_data)[grepl(names(input_data),pattern='_unique')],
            lapply(
              names(input_data)[grepl(names(input_data),pattern='_unique')],
              function(x){substr(x,1,nchar(x)-7)}
            )
          )) %>%
          rename_(.dots=setNames(
            names(input_data)[grepl(names(input_data),pattern='_sum')],
            lapply(
              names(input_data)[grepl(names(input_data),pattern='_sum')],
              function(x){substr(x,1,nchar(x)-4)}
            )
          )) %>%
          return()
      }
    },
    formArray = function(...,val,dimData=list(),metaData=list()){
      if('formArray' %in% private$.debug){
        browser()
      }
      if(missing(val)){
        if(length(list(...))==0){
          return()
        } else{
          stop("val must be supplied in order to form the incidenceArray")
        }
      }
      private$aCurrent=FALSE
      private$.dnames = NULL
      self$aDims = list(...)
      self$aVal = val
      self$aDimData = dimData
      metaDataKeys <-
        names(private$.metaData)[!(names(private$.metaData) %in% names(metaData))]
      self$metaData = c(private$.metaData[metaDataKeys],metaData)
    }
  ),
  active = list(
    frame = function(value){
      if('frame' %in% private$.debug){
        browser()
      }
      private$aCurrent = FALSE
      if(!missing(value)){
        private$.rnames = NULL
        private$.cnames = NULL
      }
      private$defaultActive('.frame','private',value)
    },
    arr = function(value){
      if('arr' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.arr)
      }
      stop("Do not write directly to the array.")
    },
    dims = function(value){
      if('dims' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.dims)
      }
      stop("Do not write directly to the dimensions.")
    },
    ndim = function(value){
      if('ndim' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.ndim)
      }
      stop("Do not write directly to the number of dimensions.")
    },
    dimData = function(value){
      if('dimData' %in% private$.debug){
       browser()
      }
      if(private$aCurrent == FALSE){
        private$updateArray()
      }
      if(missing(value)){
        return(private$.dimData)
      }
      if(is.null(value)){
        private$.dimData=NULL
      } else if(class(value) == 'list'){
        nval = length(value)
        if(nval == 0){
          private$.dimData = value
        } else if(nval <= private$.ndim){
          if(all(mapply(function(self,other){
            is.null(other) ||
            all(sapply(other,function(x){self==length(x)}))
          },
          self = private$.dims[1:nval],
          other=value
          ))){
            private$.dimData[!sapply(value,is.null)] = value[!sapply(value,is.null)]
          } else {
            stop("The dimensions don't match up.")
          }
        } else {
          stop("Invalid number of dimensions.")
        }
      } else {
        stop(paste(
          "Not sure how to make dimension metaData from object of class",
          class(value)
        ))
      }
    },
    dnames = function(value){
      if('dnames' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.dnames)
      }
      stop("Do not write directly to the dimension names.")
    },
    nrow = function(value){
      if('nrow' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.dims[1])
      }
      stop("Do not write directly to the number of rows.")
    },
    ncol = function(value){
      if('ncol' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.dims[2])
      }
      stop("Do not write directly to the number of columns.")
    },
    rnames = function(value){
      if('rnames' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.dnames[[1]])
      }
      stop("Do not write directly to the row names.")
    },
    cnames = function(value){
      if('cnames' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(private$.dnames[[2]])
      }
      stop("Do not write directly to the column names.")
    },
    colData = function(value){
      if('colData' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(self$dimData[[2]])
      }
      self$dimData[[2]] <- value
    },
    rowData = function(value){
      if('rowData' %in% private$.debug){
       browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        return(self$dimData[[1]])
      }
      self$dimData[[1]] <- value
    },
    aDims = function(value){
      if('aDims' %in% private$.debug){
        browser()
      }
      if(missing(value)){
        return(private$.aDims)
      }
      lapply(value,function(value){
        if(!all(value %in% colnames(private$.frame))){
          stop(paste(value,"is not a column of the frame"))
        }
      })
      private$.aDims = value
      private$aCurrent = FALSE
    },
    aVal = function(value){
      if('aVal' %in% private$.debug){
        browser()
      }
      if(missing(value)){
        return(private$.aVal)
      }
      if(!(value %in% colnames(private$.frame))){
        stop(paste(value,"is not a column of the frame"))
      }
      private$.aVal = value
      private$aCurrent = FALSE
    },
    aDimData = function(value){
      if('aDimData' %in% private$.debug){
        browser()
      }
      if(missing(value)){
        return(private$.aDimData)
      }
      lapply(value,function(value){
        for(val in value){
          if(!(val %in% colnames(private$.frame))){
            stop(paste(val,"is not a column of the frame"))
          }
        }
      })
      private$.aDimData = value
      private$aCurrent = FALSE
    },
    mat = function(value){
      if('mat' %in% private$.debug){
        browser()
      }
      if(missing(value)){
        if(private$aCurrent == FALSE){
          private$updateArray()
        }
        if(private$.ndim == 2){
          return(as.matrix(private$.arr))
        }
        return(apply(private$.arr,c(1,2),function(x){x[slice]}))
      }
      stop(paste(
        "Do not write directly to the mat, because it is automatically",
        "calculated.  The Observation List is called frame"
      ))
    },
    slice = function(value){
      if(missing(value)){
        return(private$.slice)
      }
      if(any(c(1,1,value) > self$dims)){
        stop("Value must be between 1 and length in that dimension.")
      }
      private$.slice = matrix(value,1)
    },
    aggregate = function(value){
      if(missing(value)){
        return(private$.aggregate)
      }
      private$aCurrent = FALSE
      if(class(value) != 'function'){
        stop(paste(
          "Not a function.  aggregate should be a function taking a single",
          "data_frame argument called input_data"
        ))
      }
      if(length(names(formals(fun=value))) != 1){
        stop(paste(
          "Not a valid function for aggregation.  A valid aggregation function",
          "must take a single data_frame argument."
        ))
      }
      private$.aggregate = value
    }
  )
)

HopkinsIDD/ForecastFramework documentation built on Nov. 10, 2019, 2:15 a.m.